An information resource flow method and system based on agent cooperation

Abstract

The application discloses an information resource flow method and system based on agent cooperation. The method acquires an information association graph, object attributes of a target object and business scene characteristics, and determines a target information demand through a demand sensing agent. According to the target information demand and information themes, information contents and association edges of information nodes, a target information gap and a directional collection task are determined. Through an information collection agent and an information analysis agent, information resources are sequentially collected and candidate graph supplement information is obtained through analysis. The candidate graph supplement information is quality checked through an information checking agent, and target graph supplement information is used to update the information association graph. Through a scene pushing agent, information units in the updated graph are reorganized in a scene, a target information package is generated and sent to the target object. The application can improve the flow efficiency of information resources.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence and information processing technology, specifically to a method and system for information resource transfer based on intelligent agent collaboration. Background Technology

[0002] Information resource circulation refers to the process of collecting, processing, organizing, and transmitting information resources from different data sources so that they can be obtained and used by the appropriate entities. As enterprises become increasingly information-driven, various information resources, such as document information, business information, and interactive information, are continuously generated during business operations. By managing and circulating these information resources in a unified manner, information support can be provided for activities such as business processing, project development, personnel training, and customer service.

[0003] However, current information resources are typically collected into databases through manual entry or periodic synchronization, and stored according to preset categories, indexes, or tags. When using these resources, users retrieve the relevant content through keyword searches or preset rules. Because this information resource flow process relies heavily on manual operation or fixed rules, frequent manual maintenance and rule adjustments are required when dealing with continuously generated information resources, resulting in low efficiency in information resource flow. Summary of the Invention

[0004] This application provides a method and system for information resource transfer based on intelligent agent collaboration, which improves the efficiency of information resource transfer.

[0005] This application provides a method for information resource transfer based on intelligent agent collaboration, including:

[0006] The system acquires an information association graph, the object attributes of the target object, and the business scenario characteristics. It then uses a demand-aware intelligent agent to determine the target information demand based on the object attributes and business scenario characteristics. The information association graph includes multiple information nodes and their associated edges. Each information node corresponds to an information unit with an information theme and information content, and each associated edge represents the semantic association between information units.

[0007] Candidate information nodes are determined based on the target information requirements and the information themes of each information node. The target information gap and the corresponding targeted collection task are determined based on the information content of the candidate information nodes and their associated edges. The target information gap includes at least one of the information content gap and the information association gap.

[0008] The information collection agent performs a targeted collection task to obtain information resources to be processed, and the information parsing agent parses the information resources to be processed into candidate map supplementary information. The candidate map supplementary information includes standardized information units and correlation relationships.

[0009] The information verification agent verifies the information quality of the supplementary information of the candidate graph based on the information association graph, obtains the supplementary information of the target graph, and updates the information association graph based on the supplementary information of the target graph.

[0010] The scenario-driven intelligent agent reorganizes the information units in the updated information association graph according to the target information needs and business scenario characteristics to obtain the target information package, and then sends the target information package to the target object.

[0011] This application also provides an information resource transfer system based on intelligent agent collaboration, including:

[0012] The demand perception module is used to acquire information association graphs, object attributes of target objects and business scenario characteristics, and determine target information demands through demand perception intelligent agents based on object attributes and business scenario characteristics. The information association graph includes multiple information nodes and their associated edges. Each information node corresponds to an information unit with information theme and information content, and each associated edge represents the semantic association between information units.

[0013] The gap analysis module is used to determine candidate information nodes based on the target information requirements and the information themes of each information node, and to determine the target information gap and the corresponding targeted collection task based on the information content of the candidate information nodes and their associated edges. The target information gap includes at least one of information content gap and information association gap.

[0014] The resource processing module is used to perform targeted collection tasks through the information collection agent to obtain information resources to be processed, and to parse the information resources to be processed into candidate map supplementary information through the information parsing agent. The candidate map supplementary information includes standardized information units and correlation relationships.

[0015] The information verification module is used to verify the information quality of the supplementary information of the candidate graph by the information verification agent based on the information association graph, obtain the supplementary information of the target graph, and update the information association graph based on the supplementary information of the target graph.

[0016] The scenario push module is used to reorganize the information units in the updated information association graph according to the target information needs and business scenario characteristics through the scenario push intelligent agent, obtain the target information package, and send the target information package to the target object.

[0017] This application also provides an electronic device, including a processor and a memory, wherein the memory stores multiple instructions; the processor loads instructions from the memory to execute steps in any of the information resource transfer methods based on intelligent agent collaboration provided in this application.

[0018] This application also provides a computer-readable storage medium storing multiple instructions adapted for loading by a processor to execute steps in any of the information resource transfer methods based on intelligent agent collaboration provided in this application.

[0019] This application also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps in any of the information resource transfer methods based on intelligent agent collaboration provided in this application.

[0020] This application's embodiments determine target information requirements by combining the object attributes and business scenario characteristics of the target object, enabling subsequent information processing to revolve around the actual needs of the target object. It locates information nodes related to the target information requirements based on an information association graph, and determines the information that needs to be supplemented based on the information content of the information nodes and their associated edges. This allows the information collection agent to acquire information resources targeting actual gaps, reducing the collection of irrelevant information. For the collected information resources, the information parsing agent converts them into standardized content that can be used for graph updates. The information verification agent performs quality verification based on the information association graph before updating the graph, reducing the impact of unsuitable information on the graph content. Based on this, the scenario push agent selects and reassembles information units adapted to the target information requirements and business scenario from the updated information association graph, and sends the resulting target information package to the target object. Thus, information requirement determination, information supplementation, graph updates, and information push can be seamlessly integrated, reducing the processing resources consumed by manual maintenance, invalid collection, and repeated searches, thereby improving the efficiency of information resource flow. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating the information resource transfer method based on intelligent agent collaboration provided in an embodiment of this application;

[0023] Figure 2 This is a schematic diagram of the target information gap determination process provided in the embodiments of this application;

[0024] Figure 3 This is a schematic diagram illustrating the process of generating supplementary information for candidate maps and updating information-related maps according to embodiments of this application;

[0025] Figure 4This is a schematic diagram of the information unit scenario-based reorganization and feedback optimization process provided in the embodiments of this application;

[0026] Figure 5 This is a schematic diagram of the structure of an information resource transfer system based on intelligent agent collaboration provided in an embodiment of this application. Detailed Implementation

[0027] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] This application provides a method and system for information resource transfer based on intelligent agent collaboration.

[0029] The information resource transfer method can be executed by an information resource transfer system, which can be integrated into an electronic device, such as a terminal or server. The terminal can be a mobile phone, tablet computer, laptop computer, or personal computer; the server can be a single server or a server cluster consisting of multiple servers. In some embodiments, the information resource transfer system can also be integrated into multiple electronic devices, with multiple electronic devices jointly implementing the information resource transfer method of this application.

[0030] In the embodiments of this application, the terms "intelligent agent," "module," or "unit" refer to a computer program or part of a computer program that has a predetermined function and works together with other related parts to achieve a predetermined goal. It can be implemented wholly or partially through software, hardware (such as processing circuitry or memory), or a combination of software and hardware. Similarly, a processor, multiple processors, or memory can be used to implement one or more intelligent agents, modules, or units. Furthermore, each intelligent agent, module, or unit can be part of a larger intelligent agent, module, or unit that includes its corresponding function.

[0031] It is understood that the specific embodiments of this application involve data related to the target object's attributes, information resource operation records, interaction context, and usage feedback information. When the embodiments of this application are applied to specific products or technologies, the collection, use, and processing of related data must comply with the laws, regulations, and standards of the relevant countries and regions.

[0032] The following sections provide detailed descriptions of each example. It should be noted that the sequence numbers of the following embodiments are not intended to limit the preferred order of the embodiments.

[0033] In this embodiment, an information resource transfer method based on agent collaboration is provided, such as... Figure 1 As shown, the specific process of this information resource transfer method based on agent collaboration can be as follows:

[0034] S101. Obtain the information association graph, the object attributes of the target object, and the business scenario characteristics. Then, through the demand-aware intelligent agent, determine the target information demand based on the object attributes and business scenario characteristics. The information association graph includes multiple information nodes and their associated edges. Each information node corresponds to an information unit with an information theme and information content. Each associated edge represents the semantic association between information units.

[0035] Among them, the information association graph is a graph data structure formed by structuring information resources from different sources. It records at least information nodes, association edges, and the correspondence between information nodes and information units.

[0036] Information nodes are graph nodes in an information association graph used to locate information units. They can include node identifiers and information unit identifiers. For example, different information nodes can locate information units corresponding to operating status, alarm events, fault causes, or processing steps.

[0037] An information unit is an information carrier configured on an information node. It may include fields such as information subject, information content, source identifier, time attribute, and version identifier, so that the source, formation time, and version of the same content subject can be saved along with the information content.

[0038] Information topics are used to characterize the content category to which an information unit belongs, and can be selected from a pre-established set of topic categories; for example, the set of topic categories may include equipment status, abnormal phenomena, causes of failure, and handling methods.

[0039] Information content is used to characterize the specific data carried by the information unit, and may include at least one of the following: text fragments, field names and field values, status values, parameter values, applicable conditions, and time information.

[0040] Relationship edges are graph structure data that connect different information nodes. They can include fields such as start node identifier, end node identifier, relationship type, relationship direction, and relationship attributes to determine what kind of relationship exists between information units.

[0041] Semantic association is used to characterize the relationship between two information units in terms of content meaning, and may include at least one of the following: membership, causality, composition, temporal sequence, reference, and similarity.

[0042] The target object is the object that receives or uses information resources. It can be a natural person object, a device object, an organization object, or a collection of objects consisting of multiple objects.

[0043] Object attributes are used to describe the relatively stable attribute characteristics of a target object or those formed over a period of observation. They may include at least one of the following: object category, role, permission scope, and resource usage.

[0044] Business scenario features are used to describe the business processing environment in which the target object is currently located, and may include at least one of the following: current task, current business process node, current processing object, and interaction context.

[0045] Target information requirements are used to characterize the information topics, information content scope, and requirement order needed by the target object in the current business scenario. They can be represented by corresponding records of target information topic identifiers, requirement matching degree, and requirement priority.

[0046] A demand-aware intelligent agent is an intelligent agent that jointly processes object attributes and business scenario characteristics to determine target information requirements.

[0047] A demand-aware agent can be specifically composed of an attribute encoding program, a contextual semantic parsing program, a feature fusion program, and a topic matching program. The attribute encoding program converts roles, resource usage, and business process nodes into corresponding features; the contextual semantic parsing program can be a keyword parsing program, an entity recognition program, an intent classification program, or a pre-trained language processing network configured with intent extraction task instructions; the topic matching program compares the fused features with the topic features of each candidate information topic. The demand-aware agent can be trained using samples composed of role attributes, resource usage features, business scenario features, and corresponding target information topics, or it can be obtained by configuring feature weights, topic keywords, matching rules, and topic matching thresholds.

[0048] For example, in the information association graph, nodes N1, N2, and N3 can be set respectively. Node N1 corresponds to the operating status information unit of device A, node N2 corresponds to the alarm information unit of device A, and node N3 corresponds to the alarm handling step information unit. The association edge between node N1 and node N2 represents the state trigger relationship, and the association edge between node N2 and node N3 represents the handling guidance relationship. When the target object O is in the device abnormal handling stage, the object attributes and business scenario characteristics can jointly point to the target information requirements such as alarm cause and handling steps.

[0049] In some embodiments, in order to enable the target information requirement to simultaneously reflect the target object's historical resource usage, current business process stage, and current interaction intent, the object attributes include role attributes and resource usage characteristics, with the resource usage characteristics determined based on the target object's information resource operation records; the business scenario characteristics include the current business process node and the interaction context, with the interaction context including the target object's current query content and the prior interaction content associated with the current query content;

[0050] Determine the target information requirements based on object attributes and business scenario characteristics, including:

[0051] Semantic parsing of the interaction context yields the intent entity of the target object;

[0052] Feature fusion is performed on role attributes, resource usage characteristics, current business process nodes, and intent entities to obtain joint features of the object scenario;

[0053] Based on the joint features of the object scenario, the matching degree of the target object's needs for each information topic is determined, and multiple target information topics and the priority of each target information topic are determined based on the matching degree of the needs for each information topic, thus obtaining the target information needs.

[0054] Among them, the role attribute is used to characterize the target object's responsibility category, operational scope, or task type in the business process.

[0055] Resource usage characteristics are used to characterize how a target object uses information resources within a preset time period. They can be composed of the number of accesses, queries, cumulative usage time, or citations corresponding to different information topics.

[0056] The preset time period is the historical data observation range used when forming resource usage characteristics. It can be configured before feature generation based on the business data update cycle, object usage frequency, or model training cycle, and can be expressed in hours, days, or number of interaction rounds.

[0057] Information resource operation records are historical operation data used to determine the characteristics of resource usage, and may include at least one of the following: access object, query term, access frequency, dwell time, download status, and citation status.

[0058] The current business process node is used to represent the current processing stage of the business task. It can be determined by both the business process identifier and the node identifier, such as an exception discovery node, a cause analysis node, or a processing execution node.

[0059] The interaction context is the contextual information consisting of the current query content and the prior interaction content associated with the current query content.

[0060] The current query content is the content corresponding to the query currently initiated by the target object, which can be query text, search conditions, interface request fields, or operation instructions.

[0061] Prior interaction content is determined from historical interaction content based on its semantic relevance or temporal proximity to the current query content.

[0062] An intent entity is used to represent the content subject that the target object is currently interested in. It may include at least one of object entities, event entities, attribute entities, and action entities, and may be represented by a combination of entity category, entity name, and entity identifier.

[0063] Object scenario joint features are features formed by fusing role attributes, resource usage features, current business process nodes, and intent entities.

[0064] Demand matching score is used to characterize the degree of demand of the target object for the corresponding information topic. It can be a matching score between 0 and 1. The larger the value, the more the corresponding information topic matches the current demand.

[0065] The preset information topic is a candidate topic configured before demand perception. Its configuration data may include topic identifier, topic name, topic keywords and topic feature representation; the demand perception agent determines the target information topic from these candidate topics.

[0066] The target information topic is an information topic determined from multiple information topics based on the degree of matching between information topics and needs.

[0067] Demand priority is used to characterize the order of demand among multiple target information topics.

[0068] Specifically, the demand-aware intelligent agent can vectorize and encode role attributes, resource usage characteristics, current business process nodes, and intent entities, and then concatenate, weighted sum, or perform attention fusion on the encoded results to obtain joint features of the object scenario. Subsequently, these joint features can be matched with the theme features corresponding to preset information themes to obtain the demand matching degree for each information theme. Demand priority can be determined according to the demand matching degree from high to low, or it can be adjusted based on the time limits corresponding to the business process nodes.

[0069] For example, for the k-th information topic, the demand matching degree can be determined according to the following formula:

[0070] ,in, ;

[0071] in, Indicates the first The degree of matching between the information topic and the needs of the user. Indicates the character attributes and the first Matching values ​​for each information topic Indicates resource usage characteristics and the first Matching values ​​for each information topic Indicates the relationship between the current business process node and the first node. Matching values ​​for each information topic Indicates the intent entity and the first Matching values ​​for each information topic This indicates the weight of the matching value corresponding to the character attribute. This indicates the weight of the matching value corresponding to the resource usage feature. This indicates the weight of the matching value corresponding to the current business process node. The weight of the matching value corresponding to the intent entity.

[0072] Each weight can be obtained through training with sample data or pre-set according to different business scenarios. The above calculation formula is only one optional implementation method and does not constitute a limitation on the method of determining the degree of requirement matching.

[0073] For example, the target object is maintenance personnel A, the current business process node is the equipment anomaly handling node, the current query content points to the temperature anomaly of equipment A, and the prior interaction content includes queries about the vibration status and lubrication status of equipment A. The demand-aware agent thus obtains the intent entities such as equipment A, temperature anomaly, vibration status, and fault handling. If α, β, γ, and δ are 0.30, 0.25, 0.25, and 0.20 respectively, for the first information topic, the matching values ​​are 0.96, 0.90, 0.94, and 0.92 respectively, then the demand matching degree of the first information topic is 0.932; for the second information topic, the matching values ​​are 0.84, 0.78, 0.82, and 0.80 respectively, then the demand matching degree of the second information topic is 0.812. The demand-aware agent can determine the first and second information topics as target information topics and set the demand priority of the first information topic to be higher than that of the second information topic.

[0074] In this way, by simultaneously utilizing the relatively stable attributes of the target object and the dynamic characteristics of the current business scenario, we can reduce the demand bias caused by determining information needs based solely on fixed labels or a single query, and provide scenario-specific demand basis for subsequent information gap identification.

[0075] S102. Based on the target information requirements and the information themes of each information node, candidate information nodes are determined, and the target information gap and corresponding targeted collection tasks are determined based on the information content of the candidate information nodes and their associated edges. The target information gap includes at least one of information content gap and information association gap.

[0076] Among them, candidate information nodes are information nodes in the information association graph that have a thematic matching relationship with the target information needs, and are used to limit the analysis scope of subsequent content gaps and association gaps.

[0077] Information content gaps are used to characterize the information content corresponding to the target information topic that does not meet the preset content requirements. This can be due to missing required fields, empty field values, or field values ​​that do not meet the requirements of data type, value range, or applicable conditions.

[0078] Preset content requirements are the content integrity requirements configured for different information themes before gap analysis. They may include the set of required fields, field data types, field value rules, and applicable conditions. For example, the preset content requirements for alarm information themes may include alarm thresholds, units of measurement, and applicable alarm conditions.

[0079] Information association gaps are used to characterize the lack of association edges between candidate information nodes and other information nodes that meet the preset relationship requirements. This can be the lack of adjacent information nodes of a specified relationship type, relationship direction, or type.

[0080] Preset relationship requirements are graph structure requirements configured for different information topics before gap analysis. They can include required relationship types, relationship directions, and allowed adjacent node types. For example, alarm information nodes can be required to have at least one associated edge pointing to fault cause information nodes and processing step information nodes.

[0081] The target information gap is the object to be supplemented from one or more content gaps and related gaps based on the target information needs, and it corresponds to the relevant target information theme.

[0082] A targeted acquisition task is used to instruct an information acquisition agent to perform information resource acquisition with a specific acquisition direction towards a target information gap. It may include task identifier, acquisition topic, gap type, execution status, and task source.

[0083] In some embodiments, in order to locate specific gaps corresponding to target information needs from two dimensions—information content and graph structure relationships—and to ensure that gap selection reflects the demand order of different target information themes, candidate information nodes are determined based on the target information needs and the information themes of each information node. Then, the target information gaps and corresponding targeted data collection tasks are determined based on the information content of the candidate information nodes and their associated edges. This includes:

[0084] Based on the topic matching degree between each target information topic and the information topic of each information node, candidate information nodes corresponding to each target information topic are determined;

[0085] Based on the information content of each candidate information node, determine the candidate information content gap corresponding to each target information topic, and based on the node degree information of each candidate information node and the distribution information of its associated edges, determine the candidate information association gap corresponding to each target information topic.

[0086] Based on the priority of the needs of each target information topic, the target information gap is determined from the gaps in the content of each candidate information and the gaps in the association of each candidate information. Targeted collection tasks are then generated based on the target information gaps. The targeted collection tasks include the collection topic and the gap type.

[0087] The topic matching degree is used to characterize the semantic similarity between the target information topic and the information topic of the information node. It can be determined by topic keyword matching, topic feature similarity or classification model output value.

[0088] The topic matching threshold is the boundary value for determining whether an information node can be used as a candidate information node. It can be set according to the recognition results of labeled topic matching samples or the expected recall range. Different thresholds can be configured for different target information topics.

[0089] The gap in candidate information content refers to the information content of a candidate information node that is missing or does not meet the requirements of the target information theme.

[0090] The preset content fields are a set of fields that need to be compared at the field level in the preset content requirements. They can record the field name, field type, whether it is required, and the allowed value range.

[0091] Node degree information is the number of associated edges connected to candidate information nodes, which can be further counted according to incoming edges, outgoing edges, or relationship types.

[0092] The distribution information of associated edges is used to characterize the distribution of associated edges in at least one dimension among relation type, association direction, and adjacency information node type.

[0093] The graph structure benchmark is a comparative benchmark for judging whether the association structure of candidate information nodes is complete. It can be formed based on the statistics of confirmed complete information nodes under the same information theme, or it can be configured according to preset relationship requirements. It can include the node degree range and the number or proportion of associated edges corresponding to each relationship type.

[0094] Candidate information gaps are relationships to be filled, determined based on node degree information and the distribution information of associated edges.

[0095] The collection topic is used to limit the information topic to which the targeted collection task is directed.

[0096] Gap type is used to indicate whether the object to be supplemented belongs to information content or related relationships.

[0097] Specifically, the similarity between the topic features corresponding to the target information topic and the topic features corresponding to the information topics of each information node can be calculated, and information nodes whose similarity reaches the topic matching threshold are identified as candidate information nodes. For each candidate information node, the content fields included in its information content can be compared with the preset content fields corresponding to the target information topic, and content that is missing or does not meet the field requirements is identified as candidate information content gaps. Node degree information can be the number of associated edges connected to the candidate information node; the distribution information of associated edges can include the number of associated edges corresponding to different relationship types, the direction of associated edges, and the type of adjacent information nodes. The node degree information and the distribution information of associated edges can be compared with the graph structure benchmark corresponding to the same information topic to determine the candidate information association gaps.

[0098] For example, the information association graph includes information nodes N1, N2, and N3. N1 corresponds to the operating status information of device A, N2 corresponds to temperature alarm information, and N3 corresponds to maintenance step information. Based on the first information theme, N1 and N2 are identified as candidate information nodes. Content comparison results show that N2 lacks applicable conditions corresponding to the alarm threshold, thus forming a candidate information content gap. N1 has a node degree of 2, while the average node degree of information nodes under the same information theme is 4. Furthermore, N1's association edges do not contain any association edges pointing to fault cause-related information nodes, thus forming a candidate information association gap. Based on the demand priority of the information theme, the above candidate gaps can be identified as target information gaps, and a targeted acquisition task with task identifier T is generated. The acquisition theme of this targeted acquisition task is the temperature anomaly handling information of device A, and the gap types include information content gaps and information association gaps.

[0099] like Figure 2 As shown, in an application scenario, the object attributes and business scenario characteristics of the target object O are transformed into target information demand Q by the demand-aware intelligent agent. In the information association graph G, N1, N2, and N3 are candidate information nodes matching the target information topic A; the information unit corresponding to N2 lacks an applicable condition field, forming an information content gap C; there is a lack of an association edge between N1 and N5 that meets the preset relationship requirements, forming an information association gap R. The gap analysis module, combined with demand priority, determines the information content gap C and the information association gap R as target information gaps, and forms a targeted collection task T accordingly. Figure 2 The static topological relationship between information nodes and associated edges illustrates the formation scenario of the target information gap, and is not used to limit the number of information nodes, the number of associated edges, or the specific information topic.

[0100] In this way, the information needs of the target object can be transformed into specific gaps relative to the existing information relationship map, so that subsequent collection is carried out around the content or relationship to be supplemented, reducing the indiscriminate collection of information resources that are irrelevant to the current business scenario.

[0101] S103. The information collection agent performs a targeted collection task to obtain information resources to be processed, and the information parsing agent parses the information resources to be processed into candidate map supplementary information. The candidate map supplementary information includes standardized information units and correlation relationships.

[0102] Among them, the information resources to be processed are the information resources obtained by the information collection intelligent agent after performing the targeted collection task. They may include at least one of document data, business data and interactive data, and may carry task identifier, source identifier, collection time and original data location.

[0103] Candidate map supplementary information is map supplementary data formed by parsing information resources to be processed and awaiting information quality verification. It includes standardized information units and the corresponding association relationships between the standardized information units.

[0104] A standardized information unit is an information unit formed according to the data structure corresponding to the information association graph, which can be configured to existing information nodes or newly created information nodes. It may include standard entity identifier, information subject, entity attribute, source identifier and time attribute.

[0105] Relationships are relational data identified from information resources to be processed and used to form or update relational edges. They may include starting standard entity identifiers, ending standard entity identifiers, relationship type, relationship direction, and relationship basis.

[0106] In some embodiments, to ensure that the collection location, scope, and method of information resources correspond to the targeted collection task and to reduce the collection of information resources outside the target information gap, a targeted collection task is performed by an information collection agent to obtain the information resources to be processed, including:

[0107] The information collection agent determines the target data source from multiple data sources based on the collection theme in the targeted collection task, and determines the target collection range based on the gap type.

[0108] Based on the data source type of the target data source, determine the target data collection method from incremental inspection, event-triggered collection, and targeted crawling;

[0109] According to the target acquisition method, information resources that match the acquisition topic and fall within the target acquisition scope are collected from the target data source to obtain information resources to be processed.

[0110] The target data source is a data source that matches the collection topic among multiple data sources, and may include at least one of file storage sources, business databases, message data sources, device data sources, and authorized network data sources.

[0111] The target collection scope is used to limit the content or relationship covered by this collection. It can include the information content fields to be collected, the time range, the object range, and the relationship type to be collected, and is obtained by converting the information content gap or information association gap corresponding to the gap type.

[0112] The data source type is used to characterize the data organization, update method, or access method of the target data source.

[0113] The target data collection method is selected based on the data source type, namely incremental inspection, event-triggered collection, or targeted crawling.

[0114] The information collection agent can specifically consist of a task parser, a data source selection program, a collection scheduler, and one or more data source connectors. Data source connectors can be file reading interfaces, database query interfaces, message subscription interfaces, device data interfaces, or authorized network access interfaces. The information collection agent can obtain its data source identifier, data source type, collectable topics, access address, access permissions, update time fields, and supported collection methods by registering these for each data source. The task parser selects the appropriate data source connector and collection scheduler based on the targeted collection task, without requiring model training as a prerequisite.

[0115] In some embodiments, in order to convert information resources of different organizational forms into information resources to be processed that can be uniformly processed by an information parsing agent, the data source connector may include at least one of a structured data connector, a semi-structured data connector, and an unstructured data connector.

[0116] The structured data connector is used to read fields and field values ​​from business databases, data tables, or business system interfaces, and convert fields from different data sources into unified field identifiers based on pre-registered field mapping relationships.

[0117] Semi-structured data connectors are used to parse log data, API return data, or hierarchical structures, field names, and field values ​​in configuration files, and convert the parsing results into a unified data hierarchy structure.

[0118] Unstructured data connectors are used to convert the format of information resources such as documents, emails, interactive text, images, or audio, extract text, or transcribe speech to obtain text content that can be used to extract entities and relationships. Format conversion, text extraction, or speech transcription can be performed by corresponding document parsing programs, text recognition programs, or speech recognition programs.

[0119] The information acquisition agent can invoke the corresponding data source connector based on the data source type of the target data source, and save the information obtained by the connector along with the task identifier, source identifier, acquisition time, original data location, and data source type to form an information resource to be processed. For structured data, the information content of the information resource to be processed may include field names and field values ​​after field mapping; for semi-structured data, the information content may include data records after hierarchical parsing; for unstructured data, the information content may include text fragments formed after format conversion or content extraction.

[0120] Each data source connector can be configured in a plug-in manner, registering the data source identifier, data source type, access interface, allowed information topics to be collected, update time field, and supported collection methods. When a new data source is added, the corresponding data source connector can be added and its configuration information registered, enabling the information collection agent to obtain the information resources from that data source without changing the subsequent parsing process.

[0121] In this way, information resources from different data sources can be formed into unprocessed information resources with unified task identifiers and source records, and the data source relationships can be preserved for subsequent information parsing, quality verification and map updates.

[0122] Specifically, incremental inspection can acquire only information resources that have changed since the last acquisition by recording the last acquisition location, update timestamp, or version identifier; event-triggered acquisition can be initiated when a new, modified, or status change event occurs in the target data source; targeted crawling can acquire information resources according to the search expression and acquisition scope corresponding to the acquisition topic. The information acquisition agent can write a task identifier T and a source identifier into the information resources to be processed, so that subsequent parsing results can be traced back to the corresponding targeted acquisition task and target data source.

[0123] For example, the subject of targeted data acquisition task T is the handling of temperature anomalies in device A. The gap types include both information content gaps and information association gaps. The information acquisition agent can determine the target data source from document data source D1 and business data source D2, identify the alarm threshold, applicable operating conditions, and occurrence time as the information content to be acquired, and identify the fault cause relationships as the association relationships to be acquired. If D1 is updated according to version releases, incremental inspection is used to obtain the newly added version document; if D2 sends a change event when an alarm record is generated, event-triggered acquisition is used to obtain the corresponding record. All acquired information resources are written into the task identifier T, thus constituting the information resources to be processed.

[0124] In this way, the information resources to be processed can correspond to the gaps in the target information in terms of source, content scope and collection time, and the source relationship of the task can be preserved for subsequent analysis.

[0125] In some embodiments, in order to convert information resources of different sources and structures into information units and relational data that can be aligned with information association graphs, the information resources to be processed are parsed into candidate graph supplementary information, including:

[0126] Preprocess the information resources to be processed to obtain the information content to be parsed;

[0127] Extract candidate entities, the information themes and entity attributes of each candidate entity, and the semantic relationships between candidate entities from the information content to be parsed. Candidate entities represent the content subjects in the information content to be parsed that can be independently identified.

[0128] Based on the semantic matching relationship between each candidate entity and its entity attributes and the information content of each information unit in the information association graph, entity alignment is performed on each candidate entity to obtain the standard entity identifier corresponding to each candidate entity.

[0129] Standardized information units are constructed based on each standard entity identifier and its corresponding information theme and entity attributes. The information content of each standardized information unit includes the corresponding standard entity identifier and entity attributes. The association relationship is determined based on each standard entity identifier and semantic relationship to obtain supplementary information for the candidate graph.

[0130] Content preprocessing is a higher-level description of making the information resources to be processed parsable, and may include at least one of data decoding, field mapping, invalid character handling, segmentation, format conversion, and content normalization.

[0131] The information content to be parsed is the information content formed after the information resource to be processed has undergone content preprocessing, and is capable of extracting entities and relationships.

[0132] Candidate entities are objects in the information content to be parsed that can be independently identified and correspond to the main body of the information content. They can include devices, components, phenomena, parameters, events, or operations, and can be recorded by the start and end positions of the entity in the information content to be parsed, the entity name, and the entity category.

[0133] Entity attributes are fields and values ​​used to describe candidate entities, such as parameter values, status values, applicable conditions, occurrence time, or source identifiers; the same candidate entity can correspond to one or more entity attributes.

[0134] Semantic relations are the content meaning relationships between candidate entities extracted from the information to be parsed. They can be represented by the candidate entity identifiers at both ends of the relation, the relation type, and the relation direction.

[0135] Entity alignment is the process of matching candidate entities and their attributes with existing information units in an information association graph to determine standard entity identifiers.

[0136] Standard entity identifiers are identifiers used in information association graphs to uniquely or relatively uniquely identify the content subject.

[0137] The entity alignment threshold is the boundary value for determining whether a candidate entity can reuse an existing standard entity identifier. It can be set according to the entity name, entity category, and the labeled matching samples corresponding to the entity attributes. Different entity categories can be configured with corresponding thresholds.

[0138] The information parsing intelligent agent can specifically consist of a content preprocessing program, an entity extraction program, a topic attribute extraction program, a relation extraction program, an entity alignment program, and a structured information generation program. The entity extraction program and relation extraction program can employ dictionary-based and rule-based processing programs, information extraction programs trained using labeled text, or pre-trained language processing networks configured with entity and relation extraction task instructions and structured output formats. Training samples can include sample text, the location and category of sample entities in the text, sample entity attributes, and sample semantic relationships. After training or configuration is completed, the corresponding program is encapsulated with the information association graph query interface and the data structure of standardized information units to obtain the information parsing intelligent agent.

[0139] Specifically, entity alignment can first recall potentially matching information units according to name, category and entity attributes, and then calculate the semantic similarity between candidate entities and potentially matching information units. When the semantic similarity reaches the entity alignment threshold, the standard entity identifier of the corresponding information unit is adopted; otherwise, a new standard entity identifier is generated for the candidate entity.

[0140] For example, the information acquisition agent obtains maintenance information corresponding to device A from the equipment maintenance data source based on task identifier T. This maintenance information includes: generating an alarm when the temperature reaches 75 degrees Celsius; checking the wear condition of component B when the vibration velocity is greater than 4.5 mm / s; and the possibility that wear of component B may cause abnormal temperature in device A. The information parsing agent extracts candidate entities such as device A, temperature alarm, and component B wear, and extracts entity attributes such as alarm threshold, vibration velocity condition, and status description, while identifying the causal semantic relationship between component B wear and temperature abnormality. After entity alignment, device A corresponds to standard entity identifier E1, and component B wear corresponds to standard entity identifier E2. Subsequently, the information parsing agent constructs standardized information units based on E1, E2, their information themes, and entity attributes, and forms association relationships based on E1, E2, and the causal semantic relationship to obtain supplementary candidate graph information corresponding to task identifier T.

[0141] In this way, information resources from different sources and with different structures can be converted into candidate map supplementary information that can be compared and updated with the existing information association map, while maintaining the correspondence between information content and association relationships.

[0142] S104. The information verification agent verifies the information quality of the supplementary information of the candidate graph based on the information association graph, obtains the supplementary information of the target graph, and updates the information association graph based on the supplementary information of the target graph.

[0143] Among them, the information quality verification is used to determine whether the supplementary information of the candidate map meets the quality requirements of the information associated with the map. The verification objects include standardized information units and their relationships. The verification results are used to control whether the corresponding information enters the map update process.

[0144] The target map supplementary information consists of standardized information units and relationships that have passed information quality verification in the candidate map supplementary information, and retains the corresponding task identifier, source identifier and verification results for map updates and change traceability.

[0145] The information verification agent can be specifically composed of a graph matching program, a time validity judgment program, a reference information acquisition program, a content difference calculation program, a conflict handling program, and a graph writing control program. The information verification agent can be obtained by configuring field comparison rules, relationship comparison rules, validity time requirements, reference data sources, conflict handling rules, and writing control conditions; it can also utilize samples labeled with duplicate states, conflict states, time validity states, and reference matching states to adjust the parameters of the corresponding classification or matching program. The graph writing control program only transmits information whose verification results meet the writing control conditions to the graph update interface.

[0146] In some embodiments, to filter candidate atlas supplementary information based on three aspects—the matching status between candidate atlas supplementary information and existing atlases, the time validity status of candidate information, and the matching status between candidate information and reference information—information quality verification is performed on candidate atlas supplementary information to obtain target atlas supplementary information, including:

[0147] Based on the information association graph, graph matching verification is performed on the supplementary information of the candidate graph to obtain graph matching information, which includes the duplication and conflict states of standardized information units and association relationships.

[0148] Based on the time attributes of standardized information units, determine the valid time information;

[0149] Obtain reference information corresponding to the standardized information unit, and determine reference matching information based on the content differences between the standardized information unit and the reference information. The reference information comes from a preset reference data source.

[0150] Based on the map matching information, time validity information, and reference matching information, the supplementary information of the candidate maps is filtered to obtain the supplementary information of the target map.

[0151] Among them, the graph matching information is used to characterize the matching status of standardized information units and associations relative to the existing information and association edges in the information association graph.

[0152] The repeating state is used to indicate whether the corresponding information content or relationship already exists, and can be represented by the state value of repeating, not repeating, or pending confirmation.

[0153] The conflict status is used to characterize whether there is inconsistency in the attribute values ​​or relationship types of the same content subject. It can be represented by a status value of conflict, no conflict, or pending review, and can record the fields or relationships that have conflicted.

[0154] The time attribute is an attribute used to reflect the time characteristics of a standardized information unit, and may include at least one of the following: generation time, release time, effective start and end time, and version time.

[0155] Time validity information is used to characterize whether the time attribute of a standardized information unit meets the valid time requirements of the corresponding information topic.

[0156] The validity period requirement is the time judgment criterion configured for different information topics before quality verification. It may include the allowed information release time range, validity period, or version relationship, and can be determined according to the update cycle and validity period of the corresponding information topic.

[0157] The effective time window is an acceptable time range formed based on the effective time requirements. Its start point, end point, or duration can be configured separately for different information topics.

[0158] A reference data source is a pre-configured data source used to provide reference information. Its configuration data may include data source identifier, accessible fields, applicable information topics, source level, and data retrieval method.

[0159] Reference information is information obtained from a pre-defined reference data source and used to compare standardized information units.

[0160] Reference matching information is used to characterize the degree of content matching between standardized information units and reference information.

[0161] Preset conflict handling rules are conflict resolution rules configured before quality verification. They can determine the information to be retained based on source level, time sequence, version sequence, or field type, and can mark conflict information that cannot meet the rule conditions as awaiting manual review.

[0162] Specifically, graph matching verification compares the standard entity identifier, information theme, and entity attributes of standardized information units with the information units in the information association graph, and compares the relationship type, starting entity identifier, and ending entity identifier of the association with existing association edges. When corresponding content or relationships already exist and the content values ​​are the same, it can be determined as a duplicate state; when the standard entity identifier and attribute type are the same but the attribute values ​​are inconsistent, or when there is a mutually exclusive relationship type between the two entities, it can be determined as a conflict state. For duplicate information, existing information can be retained or the source identifiers can be merged; for conflicting information, the information to be retained can be determined according to the source level, time attribute, or preset conflict handling rules. Conflicting information that cannot be automatically processed can be entered into the manual review queue.

[0163] Time validity information can be determined based on the generation time, publication time, validity period, or version time of the standardized information unit. For example, if the time attribute of the standardized information unit falls within the valid time window of the corresponding information topic, the time validity information can indicate validity; otherwise, it can indicate invalidity or pending review. The reference data source can be a pre-configured data source corresponding to the information topic, and the reference information can be obtained from the reference data source according to the standard entity identifier, information topic, and time range. Reference matching information can be determined based on differences in attribute values, text semantic differences, or state differences. The information verification agent can perform filtering based on graph matching information, time validity information, and reference matching information without further defining a fixed quality condition expression in the claims.

[0164] For example, the supplementary information in the candidate graph includes an alarm threshold of 75 degrees Celsius for device A, with its time attribute being the release time of the day, and its source identifier pointing to data source D1. The existing threshold corresponding to the same standard entity identifier in the information association graph is 70 degrees Celsius; therefore, the graph matching information representation is not duplicated and there are attribute value conflicts. If the effective time window for the alarm information topic is the most recent 30 days, the source level of D1 is higher than the source level of existing information, and the reference threshold for the same version in the preset reference data source D2 is 75 degrees Celsius, then the time-valid information representation is valid, and the reference matching information representation is a match; the information verification agent can retain the standardized information unit corresponding to 75 degrees Celsius according to the preset conflict handling rules. If the reference information cannot be obtained, the standardized information unit can be marked as pending review and not directly used for graph updates.

[0165] In this way, duplicate, conflicting, time-out, or mismatched content with reference information can be identified before supplementary information of candidate maps is written into the information association map, reducing the possibility of information that does not meet quality requirements entering the map.

[0166] In some embodiments, in order to configure information units that have passed information quality verification to corresponding nodes and to make the update process of nodes and associated edges traceable, the target graph supplementary information includes target standardized information units and target association relationships. Updating the information association graph according to the target graph supplementary information includes:

[0167] Based on the semantic matching relationship between each target standardized information unit and each information unit in the information association graph, the target information node corresponding to each target standardized information unit is determined. The target information node is an existing information node that matches the target standardized information unit, or a newly created information node based on the target standardized information unit.

[0168] Configure each target standardized information unit as an information unit of the corresponding target information node;

[0169] Based on the target association relationships, update the association edges between each target information node to obtain the updated information association graph;

[0170] Record the changes to each target information node and its associated edges, and generate graph version information corresponding to the information association graph.

[0171] Among them, the target standardized information unit is the information unit to be configured to the information node in the target map supplementary information.

[0172] The target association relationship is the relationship data in the supplementary information of the target graph that is to be used to add or modify the associated edges.

[0173] A target information node is the configuration position of a target standardized information unit in the information association graph. It can be an existing information node that matches the target standardized information unit, or it can be a newly created information node based on the target standardized information unit.

[0174] The node matching threshold is the boundary value for determining whether a target standardized information unit should be configured to an existing information node or a newly created information node. It can be set based on historical entity alignment results or labeled node matching samples. Different information themes can be configured with corresponding thresholds.

[0175] Change information is information that records the updates of target information nodes or related edges, and may include the change object identifier, change type, content before change, content after change, task identifier, and change time.

[0176] Graph version information is used to identify the version formed by a graph update and the corresponding changes to nodes and associated edges. It can include the current version identifier, the previous version identifier, the set of change information, and the generation time.

[0177] Specifically, for a target standardized information unit whose semantic matching degree with an existing information unit reaches the node matching threshold, the corresponding existing information node can be identified as the target information node, and the target standardized information unit can be configured to that target information node to supplement or replace the corresponding information content. For a target standardized information unit that does not reach the node matching threshold, a new information node can be created and identified as the target information node. When updating associated edges, the corresponding target information node can be found based on the starting and ending standard entity identifiers of the target association, and associated edges can be added between target information nodes or the relationship attributes of existing associated edges can be updated. Change information can include the change object identifier, change type, content before change, content after change, task identifier, source identifier, and change time. Graph version information can include the version identifier, previous version identifier, change information set, and generation time to support traceability of the update process. If a write failure or the endpoint node of the associated edge does not exist during the update process, the current update can be stopped and the previous graph version can be retained. The update can be re-executed after the exception is resolved.

[0178] For example, after verifying the supplementary information of the candidate graph corresponding to task identifier T, the temperature alarm threshold does not overlap with the existing information content in the information association graph and there are no unresolved conflicts. Its time attribute is within the valid period and matches the reference information in the reference data source. The causal relationship between component B wear and temperature anomaly does not exist in the information association graph, and the entities at both ends of the relationship can be identified. Therefore, the information verification agent determines the corresponding standardized information unit and association relationship as the supplementary information of the target graph. The target standardized information unit corresponding to device A matches the existing information node N1, so the information content of N1 is updated; the target standardized information unit corresponding to component B wear does not match the existing information node, so a new information node N4 is created; subsequently, an association edge representing the causal relationship is established between N4 and N1. The version of the information association graph is updated from V1 to V2, and the changes in the information content of N1, the new information of N4, and the new information of the association edge are recorded.

[0179] like Figure 3As shown, in an application scenario, document data source D1, business data source D2, and interactive data source D3 constitute a multi-source information environment. The information acquisition agent obtains the information resources to be processed from the corresponding data sources according to the directional acquisition task T. The information parsing agent converts the information resources to be processed into standardized information units and their relationships corresponding to standard entity identifiers E1, E2, and E3. The information verification agent determines the target graph supplementary information based on graph matching information, time validity information, and reference matching information. After the target graph supplementary information is written into the graph G(t) before the update, the information content of the existing information node N1 can be updated, and a new information node N4 and the associated edge between information nodes N1 and N4 can be created to form the updated graph G(t+1). At the same time, the change record corresponding to version V2 is generated. D1 to D3, E1 to E3, and N1 to N4 are all identifiers for ease of explanation and do not constitute a quantity limit.

[0180] Thus, by performing graph matching, time validity and reference information matching checks on the supplementary information of candidate graphs before graph updates, the possibility of duplicate, conflicting or invalid information being directly written into the graph can be reduced; by recording the change information of nodes and associated edges, the graph update process can be made traceable.

[0181] S105. The scenario-driven intelligent agent reorganizes the information units in the updated information association graph according to the target information needs and business scenario characteristics to obtain the target information package, and sends the target information package to the target object.

[0182] Contextualized reorganization is a process of filtering, sorting, and combining information units in the graph according to the current information needs and business scenarios of the target object, while retaining the correspondence between the selected information units and the associated edges during the process.

[0183] The target information packet is a set of information to be sent after being reorganized into a context. It may include an information packet identifier, a context identifier, multiple sorted information unit identifiers, information content, and relational information used to characterize the semantic relationships between these information units.

[0184] The sending method refers to the implementation form of sending the target information packet to the target object, which may include at least one of the following: interface display, message push, interface return, and task node mounting; the sending record may include the target object identifier, the information packet identifier, and the sending time.

[0185] The scene-pushing intelligent agent can be specifically composed of a graph query program, a subgraph traversal program, a scene template matching program, an information filtering program, an information sorting program, and a sending interface. The scene-pushing intelligent agent can obtain information by registering the applicable business process nodes, applicable object roles, information filtering rules, information sorting rules, and sending methods for each scene template; it can also adjust template selection parameters and sorting weights using samples including business scene characteristics and desired information arrangement results. The graph query program and subgraph traversal program are used to obtain candidate information subgraphs, and the sending interface is used to transmit target information packets to the corresponding terminals, business nodes, or interface callers according to the target object identifier.

[0186] In some embodiments, in order to filter and sort the graph information according to the current business scenario of the target object, and to retain the association relationship between information units in the target information package, the information units in the updated information association graph are reorganized in a scenario-based manner to obtain the target information package, including:

[0187] Based on the target information requirements, the target information nodes and their associated edges are determined from the updated information association graph to obtain the candidate information subgraph;

[0188] Based on the characteristics of the business scenario, a target scenario template is determined from multiple preset scenario templates. The target scenario template includes information filtering rules and information sorting rules.

[0189] Based on the information filtering rules, target information units are determined from the candidate information sub-graphs, and the target information units are sorted according to the information sorting rules.

[0190] Based on the sorted target information units and their associated edges, a target information packet is generated.

[0191] Among them, the candidate information subgraph is a set of information nodes and associated edges corresponding to the target information requirements extracted from the updated information association graph. It can be represented by seed node identifiers, node sets, and associated edge sets.

[0192] The preset scenario template is an information organization template configured before scenario-based reorganization. Its configuration data may include template identifier, applicable business process nodes, applicable object roles, applicable task types, information filtering rules, and information sorting rules.

[0193] The target scenario template is a template that matches the characteristics of the business scenario from multiple preset scenario templates.

[0194] Information filtering rules are used to limit the information units in the candidate information subgraph that can enter the target information package. These rules may include allowed information topics, relationship types, time ranges, quality statuses, and node types.

[0195] Information sorting rules are used to determine the order in which information units enter the target information packet. They may include sorting indicators, the weight of each sorting indicator, and the order determination method in case of a tie.

[0196] The target information unit is the information unit determined from the candidate information subgraph according to the information filtering rules.

[0197] The preset hop count is the maximum number of associated edges allowed to be traversed when performing graph traversal from the seed node. It can be configured according to the content range of the target information package or the association level required by the business scenario.

[0198] The preset relationship type is the set of associated edge types that are allowed during graph traversal. It can be configured according to the target information theme and business process nodes, such as causal relationship, handling guidance relationship or composition relationship.

[0199] Specifically, seed nodes can be determined in the updated information association graph based on the target information theme, and graph traversal can be performed within a preset number of hops or a preset relationship type range. The information nodes and associated edges obtained during the traversal can form a candidate information subgraph. Multiple preset scenario templates can correspond to different business process nodes, object roles, or task types. Information filtering rules can include at least one of the following: information theme range, relationship type range, time validity range, and information quality level; information sorting rules can include at least one of the following: demand priority, node relevance, time proximity, and source level. When generating the target information package, a display sequence suitable for the current scenario can be formed according to the sorting results while retaining the correspondence between information units.

[0200] For example, for maintenance personnel A at the equipment anomaly handling node, an anomaly handling scenario template can be selected as the target scenario template. The scenario push agent uses the temperature alarm information node as the seed node, and performs graph traversal along causal and handling guidance relationships within a two-hop range to obtain a candidate information subgraph including alarm threshold, fault cause, inspection conditions, and handling steps. Then, it excludes time-outdated information units according to information filtering rules, and arranges the remaining information units according to demand priority and the order of "alarm information - fault cause - inspection conditions - handling steps" to form the target information package.

[0201] In this way, the content scope and arrangement order of the target information package can be adapted to the current business scenario, and the semantic connection between information can be weakened by using the storage order of the original data source.

[0202] In some embodiments, in order to modify the requirements determination process and the scenario-based reconfiguration process by utilizing the actual usage of the target information package by the target object, after sending the target information package to the target object, the method further includes:

[0203] Obtain usage feedback information of the target object for the target information package. The usage feedback information includes at least one of the following: the target object's access status for each information unit in the target information package, usage duration, and evaluation information.

[0204] Based on the feedback information and the information themes of each information unit in the target information package, determine the feedback requirements of the target audience;

[0205] Based on the difference between the feedback requirements and the target requirements, the information adaptation deviation information is determined;

[0206] By iteratively optimizing the intelligent agent, the demand perception parameters used to determine the target information requirements and the scene reorganization parameters used to filter and arrange information units are updated based on the information adaptation deviation information.

[0207] Feedback information is used to characterize the actual use of each information unit in the target information package by the target object, and corresponding records can be established according to the target object identifier, information package identifier, and information unit identifier.

[0208] Access status is used to indicate whether the corresponding information unit has been accessed or whether the access has been completed.

[0209] Duration is used to characterize the duration for which a corresponding information unit is viewed or accessed.

[0210] Evaluation information is the assessment made by the target audience of an information unit, which may include ratings, selections, or textual feedback.

[0211] Feedback demand information is used to characterize the actual level of attention paid to each information topic by the target objects identified based on the feedback information. It can be obtained by normalizing and aggregating the access status, usage duration, and evaluation information of each information unit under the same information topic.

[0212] Information adaptation bias information is used to characterize the difference between feedback demand information and target information demand, and may include at least one of target information topic bias, demand order bias and content selection bias.

[0213] Iterative optimization agents are those that update demand perception parameters and scene reorganization parameters based on information adaptation deviation information.

[0214] Demand-aware parameters are processing parameters used to determine target information requirements, and may include model parameters of the demand matching model or fusion weights of different features.

[0215] Scene reorganization parameters are processing parameters used to filter and arrange information units, and may include information filtering thresholds, sorting weights, and template selection parameters.

[0216] The preset quantity is the cumulative number of usage feedback records required to trigger a parameter update, which can be configured according to the sample size requirements of the parameter update and the frequency of business feedback.

[0217] The preset update time is the time condition that allows startup parameter updates. It can be a fixed time point, the interval between adjacent updates, or the end time of a business cycle.

[0218] Verification conditions are the test requirements for determining whether the updated parameters can be used in subsequent processing. These can include the upper limit of the requirement matching error, the lower limit of the information filtering accuracy, or the lower limit of the ranking evaluation index on the verification data.

[0219] The iterative optimization agent can be specifically composed of a feedback aggregation program, a deviation calculation program, a parameter update program, a parameter verification program, and a parameter version management program. The feedback aggregation program summarizes and uses feedback information according to the target object identifier, information package identifier, and information topic; the parameter update program can generate candidate parameters using gradient updates, weight search, or rule-based parameter adjustments; the parameter verification program uses verification data independent of the parameter update data to determine whether the candidate parameters meet the verification conditions; the parameter version management program publishes the candidate parameters when verification is successful, and retains the parameters before the update when verification fails. The iterative optimization agent can be obtained by configuring feedback field mapping, update trigger conditions, adjustable parameter ranges, verification data, and verification conditions.

[0220] The iterative optimization agent can update the above parameters when a preset number of usage feedback messages are accumulated or a preset update time is reached, and put the updated parameters into subsequent processing after they meet the verification conditions.

[0221] In some embodiments, when the scene push agent generates a target information package, it can configure a corresponding information package identifier for the target information package and record the information unit identifier, information topic, arrangement position, and corresponding associated edges of each information unit included in the target information package. After the target information package is sent to the target object, a sending record can be established based on the target object identifier, information package identifier, and information unit identifier.

[0222] When a target object generates usage feedback information, the information resource flow system can map the access status, usage duration, or evaluation information to the corresponding information unit in the target information package based on the information package identifier and information unit identifier carried in the usage feedback information. The iterative optimization agent can aggregate the usage feedback information according to the target object, business scenario characteristics, and information theme to obtain the actual level of attention corresponding to each information theme, and compare it with the target information requirements adopted when generating the target information package to form information adaptation deviation information.

[0223] The iterative optimization agent can update the demand perception parameters based on the target information topic deviation, and update the scene reorganization parameters based on the content selection deviation and demand order deviation. The updated demand perception parameters and scene reorganization parameters can form candidate parameter versions, which are then verified using feedback records from those not involved in this parameter update. When a candidate parameter version meets the verification conditions, it is published to the demand perception agent and the scene push agent for subsequent information resource flow tasks. When a candidate parameter version does not meet the verification conditions, the parameter version before the update is retained.

[0224] The parameter versions can be recorded in relation to the information resource flow tasks that generate the target information packages, thus determining the demand-aware parameters and scenario reorganization parameters used for different target information packages. This allows the actual usage of the target information packages to be sequentially fed back to the demand-aware agent and the scenario-pushing agent, and enables the parameter update process to be verified and traced.

[0225] For example, the preset quantity can be configured to 100 feedback records, and the preset update time can be configured to the end of each business cycle. If the cumulative number of feedback records reaches 100, the iterative optimization agent adjusts the feature fusion weights and ranking weights based on the difference between the feedback demand information of each information topic and the original demand matching degree. If the demand matching error of the adjusted parameters on the verification data is not higher than that of the parameters before the update, and the information ranking evaluation index reaches the preset lower limit, the verification conditions are met and the data is put into subsequent processing; otherwise, the parameters before the update are retained. The above quantity and verification index are only used to illustrate the parameter update process and do not constitute a limitation.

[0226] like Figure 4 As shown, business scenario features are used to determine the target scenario template, including information filtering and sorting rules. The target information requirement and the updated information association graph are used to determine candidate information subgraphs. The scenario push agent uses the target scenario template to filter and sort the information units in the candidate information subgraphs, generates a target information package, and sends it to the target object. The usage feedback information generated by the target object is used to determine the feedback requirement information and information adaptation deviation information in turn. The iterative optimization agent updates the requirement perception parameters and scenario reorganization parameters accordingly, so that the updated parameters are applied to the subsequent determination of target information requirements and determination of target scenario templates.

[0227] For example, the scene push agent uses the information nodes corresponding to the first and second information topics as seed nodes to obtain candidate information subgraphs including N1, N2, N3, and N4 and their associated edges from the information association graph of version V2. After determining the target scene template based on the device anomaly handling node, it retains the information units corresponding to the alarm threshold, possible causes, inspection conditions, and handling steps according to the information filtering rules, and sorts them according to demand priority and semantic association order. The generated target information package includes the temperature alarm threshold, the causal information between component B wear and temperature anomaly, vibration speed inspection conditions, and related handling steps, and sends the association relationship to maintenance personnel A along with the target information package. If subsequent user feedback indicates that the actual attention levels of the first and second information topics are 0.960 and 0.860, respectively, while the previous demand matching degrees are 0.932 and 0.812, respectively, the corresponding information adaptation deviations are 0.028 and 0.048, respectively. The iterative optimization agent can adjust the demand perception parameters and scene reorganization parameters according to the above deviations to make the subsequently determined information demands and information sorting closer to the actual usage of the target object.

[0228] In this way, after the information association graph is supplemented and verified, the scene push intelligent agent can reorganize the scattered information units and their associations into a target information package that can be used directly according to the current business scenario. Combined with the feedback of use, the demand perception and scene reorganization are iteratively optimized, which can form a closed-loop information resource flow process from demand identification, gap discovery, targeted collection, parsing and verification, graph update to scene push.

[0229] As can be seen from the above, the embodiments of this application form target information requirements based on the object attributes and business scenario characteristics of the target object, and locate information content gaps or information association gaps in the existing information association graph based on the target information requirements, enabling the information collection agent to perform targeted collection around the gaps. The collected information resources are parsed to form standardized information units and associations adapted to the information association graph, and then used to update the information association graph after information quality verification, so that the information content and its semantic associations in the graph can be supplemented according to actual needs. Based on this, the updated information units are filtered, sorted, and combined according to the business scenario to provide the target object with a target information package adapted to the current task. Therefore, the embodiments of this application can improve the synergy between information resource supplementation, updating, and scenario-based flow, and improve the adaptability and effectiveness of information resource flow.

[0230] To better implement the above methods, this application also provides an information resource transfer system based on intelligent agent collaboration. This information resource transfer system can be integrated into one or more electronic devices. The electronic devices can be terminals, servers, etc. Terminals can include mobile phones, tablets, laptops, and personal computers, while servers can be a single server or a server cluster consisting of multiple servers. In some embodiments, the information resource transfer system can also be distributed across multiple electronic devices, with the multiple electronic devices jointly executing the information resource transfer method of this application.

[0231] In this embodiment, an intelligent agent can function as an executable processing component within a corresponding functional module, or as an independent program component invoked by the corresponding functional module. It performs corresponding information processing tasks based on the input data received by the functional module. Each intelligent agent can be deployed in the same electronic device as its corresponding functional module, or in different electronic devices, and transmits processing results through interface calls, message queues, or shared storage. The input data and processing results transmitted by different intelligent agents can carry the same task identifier to maintain the correspondence between various data items in the same information resource flow task.

[0232] For example, such as Figure 5 As shown, the information resource circulation system may include a demand perception module 501, a gap analysis module 502, a resource processing module 503, an information verification module 504, and a scenario push module 505, as detailed below:

[0233] (I) Demand perception module 501.

[0234] The demand perception module 501 is used to acquire information association graphs, object attributes of target objects and business scenario characteristics, and determine target information demands through the demand perception agent based on object attributes and business scenario characteristics. The information association graph includes multiple information nodes and their associated edges. Each information node corresponds to an information unit with information theme and information content, and each associated edge represents the semantic association between information units.

[0235] The demand perception module 501 can transmit information association graphs, object attributes, and business scenario characteristics to the demand perception agent, and receive target information requirements generated by the demand perception agent. For a detailed implementation of the demand perception agent, please refer to the relevant description in the aforementioned S101.

[0236] (II) Gap Analysis Module 502.

[0237] The gap analysis module 502 is used to determine candidate information nodes based on the target information requirements and the information themes of each information node, and to determine the target information gap and the corresponding targeted collection task based on the information content of the candidate information nodes and their associated edges. The target information gap includes at least one of information content gap and information association gap.

[0238] The gap analysis module 502 can receive the target information requirements generated by the demand perception module 501 and transmit the generated targeted acquisition task to the resource processing module 503. For a detailed description of the implementation of the gap analysis module 502, please refer to the relevant description in the aforementioned S102.

[0239] (III) Resource Processing Module 503.

[0240] The resource processing module 503 is used to perform targeted collection tasks through the information collection agent to obtain information resources to be processed, and to parse the information resources to be processed into candidate map supplementary information through the information parsing agent. The candidate map supplementary information includes standardized information units and correlation relationships.

[0241] The resource processing module 503 may include an information acquisition agent and an information parsing agent, and may also call the information acquisition agent and the information parsing agent respectively through a program interface. The resource processing module 503 can pass the targeted acquisition task to the information acquisition agent and pass the information resources to be processed obtained by the information acquisition agent to the information parsing agent, enabling the information parsing agent to convert the information resources corresponding to different data sources into candidate map supplementary information that can be used for information association map verification and updating. For a detailed implementation of the resource processing module 503, please refer to the relevant description in the aforementioned S103.

[0242] (iv) Information verification module 504.

[0243] The information verification module 504 is used to verify the information quality of the supplementary information of the candidate graph based on the information association graph by the information verification agent, obtain the supplementary information of the target graph, and update the information association graph based on the supplementary information of the target graph.

[0244] The information verification module 504 can transmit candidate graph supplementary information and information association graph to the information verification agent, and receive target graph supplementary information generated by the information verification agent. The information verification module 504 can configure standardized information units from the target graph supplementary information to existing or newly created information nodes, and update the corresponding association edges according to the association relationships in the target graph supplementary information. For a detailed implementation of the information verification module 504, please refer to the relevant description in the aforementioned S104.

[0245] (v) Scene push module 505.

[0246] The scenario push module 505 is used to reorganize the information units in the updated information association graph according to the target information requirements and business scenario characteristics through the scenario push intelligent agent, obtain the target information package, and send the target information package to the target object.

[0247] The scene push module 505 can transmit target information requirements, business scenario characteristics, and updated information association graphs to the scene push intelligent agent, and receive target information packages generated by the scene push intelligent agent. The scene push module 505 can also send the target information packages to the terminal or business processing interface corresponding to the target object through an information interaction interface, task processing interface, or message sending interface. For a detailed implementation of the scene push module 505, please refer to the relevant description in the aforementioned S105.

[0248] Each of the above modules can be implemented through an independent computer program, process, thread, or processing circuit, or multiple modules can be integrated into a single program component, with the corresponding program instructions executed by one or more processors. Specific implementation methods for each module can be found in the foregoing method embodiments, and will not be repeated here.

[0249] Therefore, in this embodiment, the demand perception module determines the target information demand of the target object in the current business scenario, the gap analysis module forms a targeted collection task corresponding to the target information demand, and the resource processing module, information verification module and scenario push module sequentially complete the information resource collection and parsing, information association map supplementation and update and target information package generation, so that the supplementation, update and scenario flow of information resources form an interconnected processing link, thereby improving the adaptability and effectiveness of information resource flow.

[0250] In some embodiments, in order to keep the processing results of different agents for the same information resource transfer task consistent, the information resource transfer system can establish a task collaboration record for each information resource transfer task.

[0251] The task collaboration record is a data record used to record the transfer of information resource flow tasks between different intelligent agents. It may include at least one of the following: task identifier, target object identifier, current processing status, previous processing result identifier, intelligent agent to be called identifier, and processing result storage location. The task identifier is used to associate the target information requirement, target information gap, targeted collection task, information resource to be processed, candidate graph supplementary information, target graph supplementary information, and target information package in the same information resource flow task.

[0252] Different intelligent agents can transmit processing results through program interfaces, message queues, or shared storage. After determining the target information requirement, the demand-aware intelligent agent can write the target information requirement and the corresponding task identifier into the task collaboration record. After the gap analysis module forms the target information gap and the targeted collection task based on the target information requirement, it updates the current processing status in the task collaboration record and enables the information collection intelligent agent to obtain the targeted collection task. After the information collection intelligent agent, information parsing intelligent agent, and information verification intelligent agent complete the information resource collection, candidate map supplementary information generation, and target map supplementary information determination in sequence, they save the processing results and corresponding task identifiers respectively. The scene push intelligent agent obtains the updated information association map, target information requirement, and business scene characteristics based on the task identifier and generates the corresponding target information package.

[0253] The current processing status can include: demand identified, gap identified, resource acquisition in progress, resource parsing in progress, information verification in progress, map updated, and information package generated. The information resource flow system can invoke the corresponding intelligent agent when the current processing status meets the start conditions for the next processing step; when any processing step fails, it can record the failure status and location, retain the previous processing result, and re-execute the corresponding processing step based on the task collaboration record after the exception is resolved.

[0254] In this way, multiple intelligent agents can continuously perform corresponding processing around the same target information needs, and reduce the mixing of intermediate data corresponding to different information resource transfer tasks.

[0255] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0256] Therefore, embodiments of this application provide a computer-readable storage medium storing a plurality of instructions that can be loaded by a processor to execute steps in any of the information resource transfer methods based on intelligent agent collaboration provided in embodiments of this application.

[0257] The storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0258] Since the instructions stored in the storage medium can execute the steps in any of the information resource transfer methods based on intelligent agent collaboration provided in the embodiments of this application, the beneficial effects that any of the information resource transfer methods based on intelligent agent collaboration provided in the embodiments of this application can achieve can be realized. For details, please refer to the previous embodiments, which will not be repeated here.

[0259] According to one aspect of this application, a computer program product or computer program is provided, comprising a computer program / instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer program / instructions from the computer-readable storage medium and executes the computer program / instructions, causing the electronic device to perform the agent-based collaborative information resource transfer method provided in the above embodiments.

[0260] The above provides a detailed description of an information resource transfer method and system based on intelligent agent collaboration provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for information resource transfer based on agent collaboration, characterized in that, include: The system acquires an information association graph, the object attributes of the target object, and the business scenario characteristics. Then, a demand-aware intelligent agent determines the target information demand based on the object attributes and the business scenario characteristics. The information association graph includes multiple information nodes and their associated edges. Each information node corresponds to an information unit with an information theme and information content. Each associated edge represents the semantic association between the information units. Candidate information nodes are determined based on the target information requirements and the information themes of each information node. Target information gaps and corresponding targeted data collection tasks are then determined based on the information content of the candidate information nodes and their associated edges. The target information gaps include at least one of information content gaps and information association gaps. The information acquisition agent executes the targeted acquisition task to obtain information resources to be processed, and the information parsing agent parses the information resources to be processed into candidate map supplementary information, which includes standardized information units and correlation relationships. The information verification agent verifies the information quality of the supplementary information of the candidate graph based on the information association graph, obtains the supplementary information of the target graph, and updates the information association graph based on the supplementary information of the target graph. The scenario-driven intelligent agent reorganizes the information units in the updated information association graph according to the target information requirements and the characteristics of the business scenario to obtain the target information package, and then sends the target information package to the target object.

2. The information resource transfer method based on intelligent agent collaboration according to claim 1, characterized in that, The object attributes include role attributes and resource usage characteristics, the resource usage characteristics being determined based on the target object's information resource operation records; the business scenario characteristics include the current business process node and the interaction context, the interaction context including the target object's current query content and prior interaction content associated with the current query content; Determining the target information requirement based on the object attributes and the business scenario characteristics includes: Semantic parsing is performed on the interaction context to obtain the intent entity of the target object; The role attributes, resource usage characteristics, current business process nodes, and intent entities are fused to obtain joint object scenario features; Based on the joint features of the object scenario, the matching degree of the target object's needs for each information topic is determined, and multiple target information topics and the demand priority of each target information topic are determined based on the matching degree of the demand for each information topic, so as to obtain the target information demand.

3. The information resource transfer method based on intelligent agent collaboration according to claim 2, characterized in that, The step of determining candidate information nodes based on the target information requirements and the information themes of each information node, and determining the target information gap and corresponding targeted collection tasks based on the information content of the candidate information nodes and their associated edges, includes: Based on the topic matching degree between each target information topic and the information topic of each information node, candidate information nodes corresponding to each target information topic are determined; Based on the information content of each candidate information node, determine the candidate information content gap corresponding to each target information topic, and based on the node degree information of each candidate information node and the distribution information of its associated edges, determine the candidate information association gap corresponding to each target information topic. Based on the priority of each target information topic, target information gaps are determined from the gaps in candidate information content and the gaps in the association of candidate information. Targeted collection tasks are then generated based on the target information gaps, and the targeted collection tasks include collection topics and gap types.

4. The information resource transfer method based on intelligent agent collaboration according to claim 3, characterized in that, The process of executing the targeted data collection task through an information collection agent to obtain information resources to be processed includes: The information collection agent determines the target data source from multiple data sources based on the collection theme in the targeted collection task, and determines the target collection range based on the gap type. Based on the data source type of the target data source, the target collection method is determined from incremental inspection, event-triggered collection, and targeted crawling; According to the target acquisition method, information resources that match the acquisition topic and belong to the target acquisition range are collected from the target data source to obtain the information resources to be processed.

5. The information resource transfer method based on intelligent agent collaboration according to claim 1, characterized in that, The step of parsing the information resource to be processed into candidate map supplementary information includes: The information resources to be processed are preprocessed to obtain the information content to be parsed; Extract candidate entities, the information topics and entity attributes of each candidate entity, and the semantic relationships between the candidate entities from the information content to be parsed. The candidate entities represent the content subjects in the information content to be parsed that can be independently identified. Based on the semantic matching relationship between each candidate entity and its entity attributes and the information content of each information unit in the information association graph, each candidate entity is aligned to obtain the standard entity identifier corresponding to each candidate entity. Standardized information units are constructed based on each standard entity identifier and its corresponding information theme and entity attributes. The information content of each standardized information unit includes the corresponding standard entity identifier and entity attributes. The association relationship is determined based on each standard entity identifier and the semantic relationship to obtain supplementary information for the candidate graph.

6. The information resource transfer method based on intelligent agent collaboration according to claim 1, characterized in that, The step of performing information quality verification on the candidate atlas supplementary information to obtain the target atlas supplementary information includes: Based on the information association graph, graph matching verification is performed on the supplementary information of the candidate graph to obtain graph matching information, which includes the duplication and conflict states of the standardized information units and the association relationships; Based on the time attribute of the standardized information unit, determine the valid time information; Obtain reference information corresponding to the standardized information unit, and determine reference matching information based on the content differences between the standardized information unit and the reference information, wherein the reference information comes from a preset reference data source; Based on the map matching information, the time validity information, and the reference matching information, the candidate map supplementary information is filtered to obtain the target map supplementary information.

7. The information resource transfer method based on intelligent agent collaboration according to claim 1, characterized in that, The step of reorganizing the information units in the updated information association graph into a scenario-based structure to obtain the target information package includes: Based on the target information requirements, target information nodes and their associated edges are determined from the updated information association graph to obtain candidate information subgraphs; Based on the characteristics of the business scenario, a target scenario template is determined from multiple preset scenario templates. The target scenario template includes information filtering rules and information sorting rules. According to the information filtering rules, target information units are determined from the candidate information sub-graphs, and the target information units are sorted according to the information sorting rules. Based on the sorted target information units and their associated edges, a target information packet is generated.

8. The information resource transfer method based on intelligent agent collaboration according to claim 1, characterized in that, After sending the target information packet to the target object, the method further includes: Obtain usage feedback information of the target object for the target information package, wherein the usage feedback information includes at least one of the target object's access status, usage duration and evaluation information for each information unit in the target information package; Based on the usage feedback information and the information themes of each information unit in the target information package, the feedback requirement information of the target object is determined; Based on the difference between the feedback demand information and the target information demand, information adaptation deviation information is determined; By iteratively optimizing the intelligent agent, the demand perception parameters used to determine the target information requirements and the scene reorganization parameters used to filter and arrange information units are updated based on the information adaptation deviation information.

9. The information resource transfer method based on intelligent agent collaboration according to claim 1, characterized in that, The target map supplementary information includes target standardized information units and target association relationships. Updating the information association map based on the target map supplementary information includes: Based on the semantic matching relationship between each target standardized information unit and each information unit in the information association graph, the target information node corresponding to each target standardized information unit is determined. The target information node is an existing information node that matches the target standardized information unit, or an information node newly created based on the target standardized information unit. Configure each target standardized information unit as an information unit of the corresponding target information node; Based on the target association relationship, update the association edges between each target information node to obtain the updated information association graph; Record the changes to each target information node and its associated edges, and generate the graph version information corresponding to the information association graph.

10. An information resource transfer system based on intelligent agent collaboration, characterized in that, include: The demand perception module is used to acquire information association graphs, object attributes of target objects and business scenario features, and determine target information demands through a demand perception agent based on the object attributes and business scenario features. The information association graph includes multiple information nodes and their associated edges. Each information node corresponds to an information unit with an information theme and information content, and each associated edge represents the semantic association between the information units. The gap analysis module is used to determine candidate information nodes based on the target information requirements and the information themes of each information node, and to determine the target information gap and the corresponding targeted collection task based on the information content of the candidate information nodes and their associated edges. The target information gap includes at least one of information content gap and information association gap. The resource processing module is used to execute the directional acquisition task through the information acquisition agent to obtain the information resources to be processed, and to parse the information resources to be processed into candidate map supplementary information through the information parsing agent. The candidate map supplementary information includes standardized information units and correlation relationships. The information verification module is used to verify the information quality of the supplementary information of the candidate graph based on the information association graph by an information verification agent, obtain the supplementary information of the target graph, and update the information association graph based on the supplementary information of the target graph. The scenario push module is used to reorganize the information units in the updated information association graph according to the target information requirements and the business scenario characteristics through the scenario push intelligent agent, obtain the target information package, and send the target information package to the target object.

Images